Advanced Micro Devices announced on Thursday that its second-generation AI servers, branded Helios, have entered full production and are scheduled to begin shipping at the end of the third quarter. The company presented the update during a keynote in San Francisco where CEO Lisa Su outlined product details and market ambitions.
The Helios servers integrate AMD's MI455X AI accelerator alongside the new Venice central processing unit, and both chips are produced by Taiwan Semiconductor Manufacturing Co. During her address, Su stated, "Helios is in full production." She added that "customer demand for Helios is extremely strong."
On stage with Su was OpenAI's top computing executive, who announced that the ChatGPT developer plans to start using Helios racks later this year. That commitment from a major AI cloud customer was presented as an early commercial endorsement of the platform.
AMD framed the Helios launch as part of a broader push to win share from Nvidia in the data center chip market, with a particular emphasis on inference workloads that handle user queries for large language model applications like ChatGPT. Su provided AMD's long-term market sizing for the computing landscape, saying the company estimates the total computing market will reach $2 trillion by 2030. Of that total, she forecasted $1.4 trillion would come from chips that accelerate AI workloads and $220 billion from central processing units. For nearer-term context, AMD valued the total computing market at $365 billion in 2025.
Su emphasized collaboration across the industry, saying, "Frankly, the only way you're going to be able to service a market like that is for us to work together as an ecosystem." The comment framed AMD's strategy as reliant on partnerships and an interoperable stack rather than a single-company solution.
The company also highlighted performance claims for its latest EPYC processor family. AMD stated that the new EPYC Venice CPU delivers 20% more performance than Nvidia's CPU, and Su said Venice is in full production and experiencing strong customer demand.
AMD noted a collaboration with chip designer Cerebras to support fast inference via the Cerebras Cloud Service. The product announcements and partnerships came alongside a market reaction: AMD shares traded roughly 3% lower on Thursday following the disclosures.
Impacted sectors: semiconductor manufacturing, data center infrastructure, cloud services and AI compute.